PersPecTIves

OpiniOn

The probability of neurotransmitter

release: variability and feedback
control at single synapses
Tiago Branco and Kevin Staras

Abstract | Information transfer at chemical synapses occurs when vesicles fuse with
the plasma membrane and release neurotransmitter. This process is stochastic and
its likelihood of occurrence is a crucial factor in the regulation of signal propagation
in neuronal networks. The reliability of neurotransmitter release can be highly
variable: experimental data from electrophysiological, molecular and imaging
studies have demonstrated that synaptic terminals can individually set their
neurotransmitter release probability dynamically through local feedback regulation.
This local tuning of transmission has important implications for current models of
single-neuron computation.
Revealing the fundamental principles of how
neural circuits are organized and operate is
a key challenge in neuroscience. In the past
60 years, intensive effort has been invested
in characterizing the properties of neuronal connections the basic substrates of
information transfer in neuronal networks.
Electrophysiological studies, as well as
light, fluorescence and electron microscopy
approaches, have advanced our understanding of the functional and morphological
characteristics of synaptic connections
between many neuron types. In particular,
two important principles have emerged from
these efforts. First, in many cases, individual
connections between two neurons are comprised of multiple synaptic contacts, the
number varying with the connection type.
Second, the synapses that mediate these connections can be functionally very different,
even when they belong to the same axon and
contact the same postsynaptic target. What
is the basis for this multiplicity and variability in the composition of a connection and
what function does it serve?
For technical reasons early investigations were limited to studies of connections
as a whole (that is, the compound output
of all participating synaptic contacts) and
to the use of indirect analytical methods

for extracting the average unitary properties of the synapses composing them. These
studies, pioneered by Katz and colleagues,
established that the strength of a connection

is fundamentally dependent on three main

factors: the number of synaptic contacts,
the size of the postsynaptic depolarization
caused by neurotransmitter release from
a single synaptic vesicle (termed quantal
size) and the probability of neurotransmitter release at each synapse1. This third
parameter, which is the main focus of this
Perspective, is termed release probability
( pr) and is a consequence of the inherently
stochastic nature of the molecular and cellular processes that drive vesicle exocytosis2.
Thus, for each action potential, neurotransmitter release has a certain likelihood of
occurrence that defines the reliability of a
synapse for transmitting the action potential
signal and determines the average synaptic
strength1.
Recently, developments in molecular and
imaging techniques have enabled studies to
go beyond the level of whole connections
and analyse functional details of the individual synaptic contacts. Using this approach,
evidence has accumulated which shows
that single terminals contributing to a connection can have release probabilities that
are diverse and that can change over time.

Box 1 | Defining the synaptic composition of a connection

Establishing the number of active neurotransmitter release sites in CNS neurons is a major technical
challenge. Why is this the case? The central problem is that unequivocally defining a synaptic
specialization between two neurons requires ultrastructural confirmation. Ideally, a complete
morphological reconstruction of the two target cells of interest and their contact points should be
carried out in ultrastructural detail, coupled with some form of physiological measurement that
establishes active neurotransmission. In reality, a full reconstruction is hard to achieve, particularly
because dendrites and axons are often long and branched. Thus, researchers have generally relied on
approaches that are less exhaustive but can still offer valuable information. One classical experimental
strategy is based on stimulating a single putative axon and recording intracellularly from target
neurons. Subsequent analysis and fitting of statistical models to the excitatory postsynaptic potential
fluctuations allows estimations of the connection parameters, including the number of release sites
a method commonly known as quantal analysis133. This method, when used in combination with
horseradish peroxidase (HRP) injections to the stimulated axon, can identify putative contact points.
A second approach relies on paired recordings and quantal analysis, combined with HRP or biocytin
filling of the recorded cells to identify contact points. This is sometimes followed by ultrastructural
analysis to confirm the presence of a synapse (FIG. 1). Until recently, this second approach has
represented the gold standard for mapping brain connectivity. Other approaches to estimate how
many synapses constitute a given connection include the use of synaptic markers, such as antibodies
directed at presynaptic proteins, in combination with some form of cell morphology visualization
for example, expression of cytoplasmic green fluorescent protein or filling the cells with fluorescent
dyes. These methods have been especially useful in cultured neurons, where the low synaptic density
facilitates identification of individual contacts. New technical advances, such as block-face electron
microscopy134, circuit tracing with engineered viruses and various genetic approaches135, already
offer major new strategies to characterize neuronal connections and will certainly revolutionize our
understanding of the details of the brain wiring diagram.

naTURE REvIEWS | NeuroscieNce

volUmE 10 | may 2009 | 373

2009 Macmillan Publishers Limited. All rights reserved

PersPectives
moreover, pr seems to be regulated with
high spatial precision. Defining the factors
that contribute to setting the efficacy of the
synaptic vesicle release process and how they
interact is crucial for understanding both the
nature and the functional value of neurotransmitter release regulation. Here we consider how individual synapses are organized
in neuronal connections and examine the
experimental evidence for heterogeneity in
neurotransmitter release between different
synaptic terminals. We focus on local feedback control as an important mechanism for
pr regulation and heterogeneity and, from a
theoretical standpoint, consider its possible
implications for information processing.
Composition of neuronal connections
as mentioned above, the number of synaptic
contacts that contribute to a connection is
one of the three main parameters that
define the function and regulation of
that connection. nonetheless, estimating this
value for a given connection is not a trivial

problem (BOX 1). The first studies to provide

definitive evidence that neuronneuron
connections in the CnS could be composed
of multiple synaptic contacts were made in
the 1980s using electrophysiological recordings combined with light and sometimes
electron microscopy. Researchers demonstrated that, in the cat spinal cord, single
axons could make almost 20 synapses onto
individual neurons of the dorsal spinocerebellar tract 3. Similarly, goldfish mauthner
cells were shown to receive up to 25 synapses
from a single interneuron4,5. Information
on connection properties has now been
accumulated for many neuron classes (FIG. 1;
TABLE 1), revealing striking variability in the
composition of connections for different cell
types. In the cerebellum, parallel fibres from
granule cells usually make only one synapse
with Purkinje cell dendritic trees6. In the
cortex, connections between pyramidal
neurons make on average five synapses7,8,
whereas interneurons usually make more
than ten synaptic contacts9,10. more extreme

examples include the climbing fibres in

the cerebellum, which establish more than
500 synapses with a single Purkinje cell11,
and the Calyx of Held nerve terminals,
where ~600 neurotransmitter release sites
are concentrated in one giant synapse12,13.
multiple and variable numbers of synapses
are also features of cultured neurons. In primary cultures of dissociated hippocampal
cells, for example, connections are typically
comprised of 5 to 20 synaptic contacts, the
number being in part determined by culture
geometry and cell density 14.
not only are connections made up of
variable numbers of synapses, but their
spatial organization can also vary greatly. at
some connections synapses are distributed
across the whole dendritic tree, but in others
they target specific regions of the postsynaptic cell. For example, synapses between
cortical pyramidal cells in the same layer
tend to be restricted to the basal and apical
oblique dendrites8, whereas if the target is an
interneuron, such as a basket cell, synapses

C
a

I
1
100 m

B
S

2/3

0.5 m

50 m
4

III

II

II
5

B
S

0.5 m

50 m

III

6
S

WM
100 m

0.5 m

50 m

Figure 1 | characterizing neuronal connections with paired recordings and morphological reconstructions. A | A light microscopy image
of a connected pair of thick-tufted layer 5 pyramidal neurons filled with
biocytin. The open circles indicate three contacts (I and II are synapses; III is
an autapse) established by the left-hand neuron. The smaller panels show
higher magnification views of these contacts. B | The same pair of neurons
after camera lucida reconstruction. The dendritic arbor of the projecting
neuron is in red and its axonal arborization is in blue. The dendritic arbor of
the target neuron is in black and its axonal arborization is in green. The

filled blue circles indicate putative synaptic contacts established by the red
and blue neuron on the black and green neuron. The inset traces show an
Nature Reviews | Neuroscience
example of a presynaptic action potential and the corresponding excitatory
postsynaptic potential, illustrating the presence of an excitatory connection. scale bar: 50 ms (both traces), 40 mv (left trace) and 800 v (right
trace). c | electron micrographs showing the ultrastructure of the synaptic
contacts indicated in part B. The arrows point to the synaptic cleft.
B, synaptic bouton; s, spine; WM, white matter. Figure is modified, with
permission, from rEF. 138 (1997) The Physiological society.

374 | may 2009 | volUmE 10

www.nature.com/reviews/neuro
2009 Macmillan Publishers Limited. All rights reserved

PersPectives
are typically made all over the dendritic
tree and the cell body 15 (see also rEF. 16).
other connections have a variety of different organizations and degrees of clustering.
Differences in synaptic spatial organization

will influence dendritic integration17,18 and will

have an impact not only on spike output but
also on the way that single synapses are regulated; the functional consequences for this
are considered later in this Perspective.

Synapses are functionally heterogeneous

Connections are typically comprised of
multiple synaptic contacts, but do individual
synapses behave similarly? Here, we address
this question by considering one of the

Table 1 | Summary of number of contacts and release probability (pr) in different connections
cNs area

Presynaptic

Postsynaptic

pr

Methods

cat spinal cord

Group 1a axons

Motor neurons

25 (rEFs 22,24,146)

01 (cB)

r21,22,24, LM22,146,
QA21,22,24

(rEFs 21,22,24)

cat spinal cord

Group 1a and 1b
axons

DscT neurons

118 (rEF. 3)

0.071 (cB) (rEF. 23)

0.060.85 (cB) (rEF. 3)

r3,23, LM3, QA3,23

Frog spinal cord

Primary afferent
fibres

Motor neurons

2172

0.150.69

r, LM, QA147

Goldfish brainstem Interneurons

Mauthner cells

328 (rEFs 4,5)

0.170.62 (rEFs 4,5)

Pr4,5, LM4,5, eM4,5,

QA4,5

cat and rat L2/3

Pyramidal cells

Interneurons

17 (rEF. 15)

00.84 (cB) (rEF. 15)

0.130.64 (rEF. 49)

Pr15,49, LM15, eM15,

QA15, caI49

cat L2/3

Interneurons

Pyramidal and spiny stellate cells 317 (rEF. 10)

ND

Pr, LM, eM10

cat and rat L2/3

Pyramidal cells

Pyramidal cells

3.90.8 (rEF. 148)

24 (rEF. 8)
7.64.7 (rEF. 149)

0.50.05 (rEF. 148)

0.460.26 (rEF. 49)
0.650.18 (rEF. 149)

Pr25,35,42,148, LM8,
QA148,149, caI49

rat L2/3

L4 spiny cells

Pyramidal cells

45 (rEF. 150)
46 (rEF. 151)

0.790.04 (rEF. 151)

Pr150,151, LM150,151,
eM151, QA151

cat and rat L4

L4 pyramidal and L4 pyramidal and spiny stellate

spiny stellate cells cells

25 (rEF. 152)
84.2 (rEF. 149)

0.690.98 (rEF. 153)

0.861.09 (rEF. 149)

Pr152, r153, LM22,153,

QA153

rat L5/6

Pyramidal cells

Pyramidal cells

28 (rEF. 7)
48 (rEF. 138)
8.14.2 (rEF. 149)

0.160.9 (rEF. 138)

0.530.22 (rEF. 149)

Pr7,138,149, LM7,138,
eM138, QA138,149

rat L5/6

Interneurons

Pyramidal cells

15 (rEF. 154)

ND

Pr, LM, eM154

rat L5/6

Pyramidal cells

Interneurons

612 (rEF. 155)

<0.1 (rEF. 155)

Pr, LM, eM155

rat cA1

Interneurons

Pyramidal cells

612 (rEF. 9)

ND

Pr, LM, eM9

rat cA1

Pyramidal cells

Pyramidal cells

2 (rEF. 156)

ND

Pr, LM, eM156

rat cA1

stratum radiatum

Pyramidal cells

318 (rEF. 157)

0.140.81 (rEF. 157)

0.060.37 (rEF. 26)

r26,157
QA157, MK26

rat cA3

Interneurons

Pyramidal cells

213 (rEF. 158)

ND

Pr, LM, eM, QA158

Guinea pig cA3

Pyramidal cells

Interneurons

13 (rEF. 159)

0.750.19 (rEF. 159)

Pr, LM, eM, QA159

rat hippocampal
cultures

excitatory cells

(Autapse)

ND

0.090.54 (rEF. 27)

0.050.9 (rEF. 28)

r27,28 MK27
FM28

rat hippocampal
cultures

excitatory cells

excitatory and inhibitory cells

319 (rEF. 14)

0.030.9 (rEF. 14)

Pr, FM, eM, QA14

rat cerebellum

climbing fibres

Purkinje cells

51050 (rEF. 25)

221392 (rEF. 11)

0.90.03 (rEF. 25)

r25, QA25, LM11, eM11

rat cerebellum

Parallel fibres

Purkinje cells

12 (rEF. 6)

0.05 (rEF. 160)

r160, M160, LM6, eM6

rat cerebellum

Interneurons

stellate and basket cells

ND

0.10.54 (rEF. 161)

r, Ms161

striatum

L4/5 afferents

Medium spiny neurons

ND

0.42 (rEF. 162)

r, QA162

rat auditory
brainstem

calyx of Held

Principal cells in MNTB

637113 (rEF. 13)

0.250.4 (rEF. 13)

Pr, QA13,163,164

striatum

Thalamic
afferents

Medium spiny neurons

ND

0.72 (rEF. 162)

r, QA162

Olfactory bulb

Olfactory
receptor neurons

Principal mitral and tufted cells

ND
and periglomerular interneurons

0.920.03 (rEF. 6)

r, QA6

Olfactory bulb

Interneurons

Juxtaglomerular cell

0.210.32 (rEF. 165)

r, Ms165

ND

summary of the number of release sites and pr for some connections in the cNs, illustrating the diversity of these parameters across different connections. cA1,
hippocampal area cA1; cA3, hippocampal area cA3; caI, ca2+ imaging; cB, compound binomial; DscT, dorsal spinocerebellar tract; eM, electron microscopy; FM, FM-dye
based method; L, layer; LM, light microscopy; M, modelling; MK, MK-801 method; MNTB, medial nucleus of the trapezoid body; Ms, minimal stimulation; N, number of
contacts; ND, not determined (no absolute value was estimated); Pr, paired electrophysiological recording; QA, quantal analysis; r, electrophysiological recording.

naTURE REvIEWS | NeuroscieNce

volUmE 10 | may 2009 | 375

2009 Macmillan Publishers Limited. All rights reserved

PersPectives
crucial determinants of synaptic performance, pr. This variable not only defines the
reliability of synaptic transmission, but also
changes with the short-term activity history of the synapse, thus shaping the way
in which a connection dynamically adapts
to input 19,20. although pr is a fundamental
synaptic parameter, it is difficult to measure
directly and many investigations rely on
methods that can provide only estimates of
its value (BOX 2).
In most studies of synaptic function, pr is
considered to be the same for all terminals in
a connection between two neurons and the
measured pr therefore represents the average
of all contributing synapses. as such, this
value provides no specific information about
the properties of individual terminals,
which could in fact be highly variable, as
del Castillo and Katz insightfully noted when
they first formulated their theory of synaptic
transmission1. Indeed, early experiments
using classic quantal analysis of postsynaptic
responses in a single cell suggested that there
is considerable variability across terminals
of the same axon. For example, response

amplitude histograms from recordings of

cat spinal cord neurons were shown to be
better fit by a compound binomial model,
in which each release site has a different pr ,
than by a simple binomial model2123. Similar
conclusions were drawn from variance
mean analysis applied to both spinal cord
neurons24 and cerebellar climbing fibres25.
also, in Ca1 pyramidal cells, nmDa
(N-methyl-d-aspartate) current block curves
(BOX 2) were better fit with a bi-exponential
curve than with a simple exponential curve,
implying that the connections had at least
two groups of synapses with very different
prs26. Even in autaptic cell cultures, in which
a cell makes synaptic contacts with itself, pr
seems to be extremely non-uniform across
different boutons, as demonstrated both by
the nmDa current block approach27 and
by fluorescence-based measurements of pr
at individual synapses28. Based on the latter
methodology, murthy et al. reported a wide
and continuous distribution of prs across
different boutons, skewed to larger values
and with a coefficient of variation larger
than 0.5 (rEF. 28), a finding that was recently

Box 2 | Estimating release probability

Release probability (pr ) can be estimated in many different ways. Some methods allow only relative
comparisons whereas others provide absolute pr estimates for either connections as a whole or
individual synapses.

Quantal analysis
Analysis and binomial-model fitting of synaptic-response amplitude fluctuations is the classic
method for extracting quantal parameters, including pr. Many techniques for quantal analysis are
available, and all of them require long and stable electrophysiological recordings133.
Paired-pulse ratio
The degree of facilitation or depression of a synaptic connection depends on pr and can be quantified
by the paired-pulse ratio (PPR), which is defined as the amplitude ratio of the second to the first
postsynaptic response after stimulating the connection with two action potentials50. An important
caveat of this method is that the relationship between pr and PPR is not known in most cases28,136.
Failure rate
The frequency of failures of the synaptic connection can be used as an indication of pr137,138. A
major problem is that the failure rate depends on both pr and the number of release sites, and so
differences in the number of failures can be due to either of these parameters. An alternative is to
stimulate single synapses, by careful placement of an electrode directly adjacent to a fluorescently
labelled synapse139,140.
Progressive block of NMDA synaptic current
When NMDAR (N-methyl-d-aspartate receptor)-mediated synaptic currents are recorded in the
presence of an irreversible open-channel blocker (MK-801), the response amplitude is
progressively reduced owing to the increasing number of receptors that become blocked after
use. The rate of the block depends on how often glutamate is released, and a kinetic model can be
fitted to the block curve and used to convert it into pr26,27.
optical methods
FM dyes14,28,141 and vesicle proteins tagged with pHluorins (recombinant fluorescent pH
indicators)30,142 allow pr measurements at single synapses by imaging and quantifying vesicle
exocytosis. When the signal/noise ratio is high enough to allow detection of single fusion
events142,143, pr can be directly measured. Otherwise, estimations can be obtained from the average
release in response to a series of action potentials. A different approach, termed optical quantal
analysis, detects release events by imaging postsynaptic NMDAR-mediated Ca2+ accumulation in
single contacts49,140,144,145.

376 | may 2009 | volUmE 10

confirmed by other groups14,29,30 (FIG. 2a,b).

moreover, morphological correlates of pr at
synapses, such as the active zone size and the
number of docked vesicles, have been shown
to exhibit a similar distribution31.
additional support for the idea that
synapses from a single axon can have different prs comes from work that investigated
the synaptic properties along one axon that
contacted different cell types. The first evidence for heterogeneity among synapses that
share an axon but have different targets came
from studies on the neuromuscular junction
(nmJ). Direct focal recordings from different regions of an endplate belonging to one
motor neuron axon, in the crayfish opener
muscle, showed that boutons at different
locations have different prs3235. It was also
demonstrated that an axon that innervates
two different types of muscle can exhibit
facilitation at one target and depression at the
other 36. Similar findings were reported from
focal recordings at the frog nmJ3739 and
from double recordings from two muscles
contacted by the same axon in the lobster
stomatogastric system40. This type of experiment, in which the short-term plasticity of
synapses belonging to one axon is measured
in two or more postsynaptic targets, has also
been widely used to reveal pr variability in
the CnS. In the leech nociceptive system two
different motor cells can be innervated by the
same sensory cell, and the two synapses display opposite forms of short-term plasticity 41;
the same phenomenon has been observed
in the cat spinal cord42, in the giant reticulospinal axon synapses onto spinal neurons
in the lamprey 43, in hippocampal cultures44
and in cortical layer 2/3 and 5 circuits45,46.
For some multiple-target recordings in
cricket 47 and locust interneurons48, quantal
analysis confirmed that different prs underlie the differences in short-term plasticity.
more recently, in a remarkable tour de force,
Koester et al. performed optical quantal
analysis using single-synapse Ca2+ imaging
in layer 2/3 pyramidal cells and interneurons, and directly showed that pr varies with
the postsynaptic cell type49 (FIG. 2c).
In most studies that demonstrated pr variability between synapses, pr was measured
in a resting (basal) state, in response to a
single action potential. However, as mentioned above, pr is dynamic and varies on a
very short timescale50. Therefore, although
heterogeneity of basal prs is evident, during continuous activity differences between
synapses will also be strongly influenced by
the size and replenishment rate of vesicle
pools51. as such, the basal pr values might be
the most influential neurotransmitter release
www.nature.com/reviews/neuro

2009 Macmillan Publishers Limited. All rights reserved

PersPectives

Feedback control of pr
The marked heterogeneity of pr for synapses
along an axon raises the question of what
actually determines pr at a single synapse.
neurotransmitter release is the final result
of a complex series of cellular and molecular
steps, in which an action potential increases
the intracellular Ca2+ concentration and
triggers full or transient fusion of synaptic
vesicles with the plasma membrane2,52. The
success of this whole process fundamentally
depends on three variables: the number of
release-ready vesicles, the Ca2+ concentration
in the presynaptic terminal and the molecular coupling between Ca2+ and vesicle fusion.
a large number of factors can modify these
variables, including the regulation of Ca2+
channel function53, the modulation of release
machinery proteins2, the regulation of Ca2+
entry by the action potential waveform at
the presynaptic terminal54 and by different
Ca2+ buffering capabilities55, and the different architectures of Ca2+ channels and
sensors56,57 (FIG. 3). These and many other
examples, the detailed description of which
is beyond the scope of this article, clearly
show that pr regulation is a complex process,
and that the pr of a synapse depends on the
balance of all these factors. But what engages
these pr regulators? For example, different
synapses from the same axon in hippocampal cultures have different distributions of
Ca2+ channel subtypes58,59. So what determines the types of Ca2+ channels expressed
at the terminal or the phosphorylation state
of a particular release machinery protein?
The answer is not entirely clear, but some
pr regulators are determined by the nature
of the cell itself and its developmental programme, whereas others depend on the cells
environment and network activity. In recent
years, feedback from the postsynaptic site
has emerged as a major influence on the
variables and regulators that determine pr. In
this section we discuss examples that illustrate how the postsynaptic site can contribute to the determination of pr at individual
synaptic contacts.

Postsynaptic cell identity. as previously

mentioned, there are many examples of
one axon that contacts different targets

120

600 APs (1 Hz)

100
80
60
40
20
0

15 m

8 10 12

Time (min)

Number of observations

Fluorescence (%)

parameter in cells that fire infrequently, but

other factors should be taken into consideration for neurons with high output rates. For
example, although marked heterogeneity
in release is apparent during low-frequency
stimulation in hippocampal neurons in culture, after 60 action potentials at 10 Hz most
terminals release at similar rates51.

4
3
2
1
0
0.0 0.2 0.4 0.6 0.8 1.0

Release probability

50 m

50 m
50 mV
200 ms
Vm

50 mV
200 ms
Vm

F/F 1.0

5 m F/F

200 ms

F/F 1.0

5 m F/F

200 ms

Figure 2 | Variability of release probability measured at single synapses. a | An epifluorescence

image of a connected pair of hippocampal neurons in culture, with presynaptic terminals labelled with
Reviews
FM4-64 (red). The presynaptic cell is yellow and the target cell is blue. TheNature
inset plot
shows| Neuroscience
the recorded
action potential (AP) and the corresponding excitatory postsynaptic current. b | FM dye destaining
curves of the synapses between the two cells (left plot) show considerable heterogeneity, reflecting
the broad distribution of release probabilities for this connection (right plot). c | Two-photon images of
pyramidal cells (yellow), filled with a ca2+ indicator, connected to (left panel) a bitufted interneuron
(blue) and (right panel) a multipolar interneuron (blue). The presynaptic ca2+ signal in response to an
action potential was measured at single synaptic contact points between the cell pairs (boxed in the
upper panels and magnified in the lower panels). ca2+ transients are bigger when the pyramidal cell
target is a multipolar neuron, indicating that there are target-specific differences in release probability.
F/F, relative ca2+ fluorescence change; vm, membrane voltage. Parts a and b are modified, with permission, from rEF. 14 (2008) cell Press. Part c is reproduced, with permission, from rEF. 49 (2005)
American Association for the Advancement of science.

exhibiting different release properties, suggesting that the identity of the postsynaptic
cell is an important determinant of pr. This
postsynaptic influence can in principle be
exerted either during synaptogenesis or
through retrograde regulation after synapse

naTURE REvIEWS | NeuroscieNce

formation. Furthermore, the fact that synapses next to each other in a single axon can
contact different cells60 suggests that this
type of pr regulation can be restricted to
single boutons. one notable example
of such compartmentalized differentiation of
volUmE 10 | may 2009 | 377

2009 Macmillan Publishers Limited. All rights reserved

PersPectives
presynaptic properties is the demonstration
that terminals of hippocampal Ca3 pyramidal cells that contact metabotropic glutamate
receptor 1 (mGluR1)-positive interneurons have ten times as much mGluR7 as
terminals that contact mGluR1-negative
cells61. This specificity is so high that even
active zones that belong to the same terminal
but contact different targets exhibit these
differences61. This biochemical divergence
translates to a functional one, as the activation of presynaptic mGluR7 has been shown
to depress neurotransmitter release62.
Long-term Hebbian plasticity. Release probability is modulated by long-term synaptic
plasticity. In the hippocampus, for example,
high-frequency stimulation of mossy fibres
causes long-term potentiation in pyramidal cells, owing to increased vesicle fusion
efficiency following activation of the cyclic
amPprotein kinase a cascade and RIm1
(also known as RImS1) phosphorylation
in the presynaptic cell63. Interestingly, pr
Presynaptic
terminal
Ca2+
channels

Vesicle
trafficking

Vesicle
turnover
Synaptic
proteins

CAMs

Retrograde messenger, CAMs

Dendrite
Development
LTP and LTD
Short-term plasticity
Homeostatic plasticity

Figure 3 | Postsynaptic influences on release

Nature
Reviews
| Neuroscience
probability. several
processes,
such
as developmental changes and synaptic plasticity, can be
initiated in the postsynaptic terminal and retrogradely modulate release probability through
various targets in the presynaptic terminal. These
targets include factors that change the number
of vesicles available for release (by affecting vesicle trafficking or turnover), factors that change
the amount of ca2+ that enters the presynaptic
terminal (for example, by influencing ca2+ channels), factors that change the nature and properties of the synaptic proteins that form the release
machinery, and factors that change the way in
which these and other variables interact.
Feedback communication between the two sides
of the synapse can be mediated by a secreted factor or can operate directly through, for example,
cell adhesion molecules (cAMs). LTD, long-term
depression; LTP, long-term potentiation.

modulation in this context is also exquisitely

dependent on the nature of the postsynaptic
cell: at mossy fibreinterneuron synapses the
same protocol causes a long-term decrease
in pr as a result of presynaptic mGluR7
activation and protein kinase C-dependent
inhibition of voltage-gated Ca2+ channels
at the terminal64,65. This process requires
elevation of Ca2+ in the postsynaptic cell,
again indicating retrograde regulation of
pr 66. although in this example the nature of
the retrograde messenger is not clear, in several brain areas6772 long-term pr depression
results from endocannabinoid release from
the dendrite. Endocannabinoids activate
presynaptic cannabinoid receptors, leading
to inhibition of voltage-gated Ca2+ channels
and activation of K+ channels (resulting in
terminal hyperpolarization), which ultimately decreases pr 73. although endocannabinoids are diffusible messengers, studies
in the hippocampus suggest that their effects
are locally restricted73. This would make
them highly appropriate for implementing pr
control in a local and dynamic manner.
Synaptic homeostasis. other examples of
pr regulation come from studies of synaptic
homeostasis. Despite initial controversy,
it has now been shown that pr can also be
modified by this form of plasticity. This
was first convincingly demonstrated in the
Drosophila melanogaster nmJ, in elegant
experiments carried out by Davis and colleagues. Their studies showed that when
a motor neuron is biased to differentially
innervate two adjacent muscle targets, both
targets nevertheless develop normal levels of
depolarization, in part owing to homeostatic
adaptations in presynaptic neurotransmitter
release and active zone density 74,75. also, in
mutants in which postsynaptic excitability
was decreased74,7678, pr increased to restore
normal levels of activity; this process was in
part mediated by changes in voltage-gated
Ca2+ channels79,80 and active zone structure81.
Homeostatic plasticity can also change pr
in hippocampal cell cultures. Blocking glutamate receptors or preventing action potential generation leads to an increase in the
frequency of miniature excitatory postsynaptic
currents and overall pr 8284. This is associated
with increases in the sizes of the total and the
recycling vesicle pools85, suggesting that scaling of vesicle pools for example, through
modulation of vesicle trafficking 86,87
could be an efficient means of regulating pr.
Homeostatic regulation of pr can also result
from changes in synaptic vesicle recycling,
active zone size and the number of docked
vesicles85. Similar presynaptic adaptations are

378 | may 2009 | volUmE 10

seen in hippocampal organotypic slice cultures88 and cortical cultures89. Recently, it was
shown that in dissociated hippocampal cultures these changes can be synapse-specific
and are triggered by dendritic depolarization14. Furthermore, the same study showed
that pr homeostatically adapts to the synaptic
density of each dendritic branch: the more
synapses one axon makes on a dendritic
branch, the lower the pr of each synapse.
Interestingly, quantal analysis of paired
recordings of l2/3 pyramidal cells showed
an inverse relationship between pr and the
number of contacts in the connection90.
again, this argues for a feedback control of pr.
although the nature of the retrograde
messenger in homeostatic pr changes is
mostly not known, work in the D. melanogaster nmJ suggests that growth factor
signalling pathways might be involved81,91.
In principle, however, a number of different
mechanisms could have important roles,
including modulation of the release machinery by cell adhesion molecules such as postsynaptic density 95 (PSD95)neuroligin92
(by trans-synaptic activation of signalling
cascades) or any other messengers that are
involved in long-term potentiation and
depression.
Short-term activity. Release probaility also
changes with the short-term history of synaptic activity. For example, prolonged stimulation or depolarization of the postsynaptic
target can suppress neurotransmitter
release93 through an endocannabinoiddependent feedback loop, which also shows
target cell heterogeneity. once again, this
emphasizes pr control by the postsynaptic
cell94,95. The most classic forms of short-term
plasticity, however, occur over a shorter
timescale, and their dynamics depend on a
number of properties of the terminal50; these
include the vesicle pool size, the vesicle recycling rate, the level of Ca2+ buffering and the
expression of different kinds of metabotropic
and ionotropic receptors. Interestingly, many
of these properties have been shown to vary
from synapse to synapse, indicating that not
only the resting pr but also the dynamic pr
response to activity can be regulated at the
level of single synapses96101 (reviewed in
rEFs 102,103).
In summary, pr is influenced by a large
range of factors acting through various
mechanisms and targets. although some pr
adjustments can be induced by the presynaptic terminal itself 104, most rely on a feedback
loop from the dendritic target, suggesting
that pr is highly controlled by the identity
and activity of the postynaptic cell. although
www.nature.com/reviews/neuro

2009 Macmillan Publishers Limited. All rights reserved

PersPectives
pr is also likely to be affected by cell-wide
adjustments of synaptic function105, a substantial body of evidence suggests that pr
changes can be tailored to the specific needs
of individual synapses. at any given point
in time, therefore, the pr of a synapse will be
the result of all of these influences, each of
which has a different weight, timescale and
duration of action. In view of this, it perhaps
is less than surprising that any two terminals
belonging to the same axon can exhibit
strikingly different prs.
Functional implications
It is evident that many neuronal connections
are composed of multiple unreliable synaptic
contact points, and that the probability of
successful neurotransmitter release is variable and adjustable at the single-synapse
level. Why is this so? Is there any functional
advantage to this design?
The consequences of synaptic unreliability for information transfer have been widely
debated106111. There is a general consensus
that one of the most important outcomes
of having synaptic prs <1 is the flexibility it
provides. The short-term activity dependence of pr means that synaptic strength is
dynamic, and that synapses can act as filters
of the input pattern of the presynaptic action
potential20,112114. an adjustable pr therefore gives synapses a broad dynamic range
that is sensitive to the activity pattern, and
can function as a gain control mechanism.
Furthermore, setting the basal pr to <1
also permits longer-term plasticity-driven
changes in synaptic weight an effective
means of changing the strength and dynamics of a connection115. Given that sometimes
stimulation does not result in synaptic
transmission at a contact point (that is,
there is synaptic failure), it seems intuitively
desirable that connections be composed
of more than one contact point to ensure
that information transfer occurs every
time116. also, a high number of release sites
per connection with pr <1 would seem to
permit high fidelity of transmission during
prolonged high-frequency stimulation. at
steady state, when the readily releasable pool
of vesicles at a synapse has been depleted
and the maximum rate of release depends on
the kinetics of vesicle pool replenishment,
having multiple synapses may offer a further
advantage: while presynaptic terminals that
have recently undergone vesicle fusion are
in recovery and therefore not available for
transmission, other synapses in the connection will be operational. Thus, a more
constant overall release rate can be achieved.
nevertheless, other specializations such

Excellent S/N (5:1)

High energy use
Signal

Good S/N (3:1)

Medium energy use

Poor S/N (1:1)

Low energy use

Noise

Local pr
compensation

Global pr
compensation

Noise 1
decreases
Signal
1

Noise 1

2
S/N maintained
Less energy use

S/N maintained

S/N maintained
Less energy use

S/N degraded

S/N recovered

S/N maintained

S/N recovered

S/N increased
More energy use

Noise 2

Noise 1
increases

Figure 4 | consequences of pr adjustments for signal/noise ratio and energy usage. a | A neuron
receiving an input of interest (signal) and noise. Filled circles at the endNature
of each
input are
active synReviews
| Neuroscience
apses and open circles are inactive synapses a representation of release probability (pr ). In the three
examples shown the pr of the noise is constant and the pr of the signal varies, leading to different signal/noise (s/N) ratios and energy expenditures. The optimal arrangement is the middle one, in which
an adequate s/N ratio is achieved with a minimum number of active synapses. b | The same schematic
representation as in part a, but in this case two independent sources of noise arrive at two different
dendrites. The response to changes in the noise of dendrite 1 is illustrated using two possible modes
of pr regulation local, in which pr adapts only in synapses on dendrite 1, and global, in which pr
changes are made for all synapses of the signal. Local pr changes lead to energy-efficient s/N ratio
maintenance, whereas global pr regulation may result in s/N ratio degradation or in unnecessarily high,
energetically expensive s/N ratios.

as large vesicle pools, fast vesicle recycling

rates117,118 and neurotransmitter release without full vesicle collapse119 are necessary for
synapses that operate at high frequencies.
It also makes sense from an energyefficiency standpoint to ensure that information about action potential firing is transmitted every time, given that action potential
generation is a highly energy-consuming
process120,121. on the other hand, recovery
from the postsynaptic actions of neurotransmission is also very energetically costly 120,121,
and so for maximum efficiency the number
of active synapses per connection should be
kept to the minimum that ensures a signal
above noise122124. an adjustable probability
of neurotransmitter release under feedback
control seems a sensible mechanism by
which to constantly and quickly adapt to

naTURE REvIEWS | NeuroscieNce

the postsynaptic noise level while ensuring

minimal energy consumption (FIG. 4a).
If one assumes that the goal of synaptic
transmission between two neurons is for the
presynaptic cell to influence the firing
of the postsynaptic cell, in principle feedback
control of pr as a means of maintaining efficient signal transfer should be implemented
in a cell-wide manner, so that it acts on all
synapses that compose a connection. If synaptic signals were linearly transmitted from
the dendrites to the soma this would be an
efficient means of control because only voltage fluctuations near to the action potential
initiation site would influence the signal/
noise ratio. However, dendritic branches
are very independent electrotonic
compartments125127 that can be highly
nonlinear owing to the presence of several
volUmE 10 | may 2009 | 379

2009 Macmillan Publishers Limited. All rights reserved

PersPectives
voltage-gated conductances128 and to
shunting129. Thus, local voltage fluctuations
have an impact on the degree of nonlinearity
(by changing the recruitment of voltagegated channels and the degree of shunting)
and change the magnitude of signals that
reach the action potential initiation site. one
can therefore propose that feedback adjustments of pr that maintain an adequate signal/
noise ratio are implemented by the cell at the
compartment level to account for the local
noise (FIG. 4b). The observation that synapses from one axon onto a single dendritic
branch have more similar prs than synapses
onto different branches is consistent with
this hypothesis14. Interestingly, in most connections that have been documented to date,
synaptic terminals in a single connection
tend to contact different dendritic branches.
The reason for this is unclear, but it could
be a strategy to reduce shunting and to provide a more complete sampling of the dendritic tree and the overall activity of the cell.
another potential reason for the local
regulation of pr is that neuronal compartments might perform regional integration
operations, acting as semi-independent
computation units18,130,131. In this scenario,
in which a neuron can be thought of as a
multiple-unit network, it makes sense that
Glossary
Active zone
A specialized area of the presynaptic membrane where
synaptic vesicle exocytosis occurs.

Dendritic integration
The process through which synaptic inputs interact with
each other and with the electrical properties of the
dendritic tree to generate patterns of action potential
output.

Gain control
regulation of the relationship between synaptic input and
neuronal output.

Miniature excitatory postsynaptic currents

The postsynaptic signals that are produced in response to
spontaneous release of a single quantum of transmitter
(usually a single vesicle).

Quantal analysis
A statistical approach for decomposing the synaptic
response into the underlying quanta, usually involving
estimation of quantal size, release probability and number
of release sites.

Shunting
A decrease in the size of synaptic responses that results
from an increase in the membrane conductance (for
example, through neurotransmitter-activated ion
channels).

Synaptic homeostasis
A regulatory process that stabilizes synaptic weights
around a set point.

signal/noise adjustments are performed

separately for each unit rather than for the
cell as a whole. also, having synapses with
different prs in different dendritic branches
means that information from a single axon
can be dynamically filtered in a different way
at each dendritic compartment. another
argument for neurons to use single-synapse
pr adjustments is that a more local regulation of synaptic weights might be easier to
implement than a global one. although a
globally coordinated adjustment of pr could
be achieved by a glia-secreted factor such
as tumour necrosis factor-132, such a uniform adjustment of synapses in all dendritic
branches would require coordinated release
from multiple glial cells, adding extra complexity to the regulation process. By contrast,
local retrograde messengers or cell adhesion
molecules can efficiently reach their target
with high precision and specificity.
local feedback regulation of pr provides
an explanation for pr heterogeneity in a connection. However, it is important to note that
it does not imply that each synapse should
necessarily be different. Terminals belonging
to the same axon generate similar levels of
postsynaptic activity because they have the
same presynaptic firing history. If the local
dendritic environment of each terminal is
also similar for example, owing to functional spatial segregation of inputs on the
dendritic tree then pr should be essentially
uniform among synaptic contacts in a given
connection, as has been found for some
cortical connections49,90.
We think that the model of local pr regulation that we discuss here should be taken
only as a general principle that particular
cells and circuits tailor to suit their functional
needs. For example, a single climbing fibre
makes more than 500 synapses with a pr >0.9
onto a Purkinje cell, whereas axons from the
Schaffer collaterals in the hippocampus make
a small number of very unreliable synapses
onto Ca1 pyramidal cells. The reason for
this diversity most likely resides in the type
and importance of the information carried
by these fibres, as well as in the degree of
circuit redundancy and plasticity. a variable
probability of neurotransmitter release that is
independently adjustable for each single synapse seems to be an excellent means by which
to ensure optimal functional specialization
for different connections.

implications. But what is the role of this heterogeneity? Does it merely relate to noise in
the system and is it simply a consequence of a
non-optimized design? or does it reflect the
particular history and environment of each
synapse and influence synaptic function in
ways that can change the output of neuronal
networks? The answers to these questions are
unclear at present and will probably emerge
only from detailed theoretical models of connections between neurons in combination
with experiments designed to better understand synaptic integration in complex dendritic trees. additionally, many aspects of pr
regulation need to be explored before definitive answers can be reached. For example, the
nature of the regulators and effectors must be
identified through experiments that manipulate candidate proteins and pathways, so that
we can begin to understand how different
influences such as long-term potentiation
and depression and homeostatic plasticity
interact to shape pr. Furthermore, the exact
spatial precision of pr regulation needs to be
characterized and the monitoring variables
that trigger pr changes need to be identified.
are adjustments in pr driven by the local Ca2+
level in the dendrite, for example? If so, what
is it that determines whether a cell triggers a
Hebbian-like change or a homeostatic one?
Finally, although much information can be
gathered from experiments in cell culture
and brain slices, links between the details
of synaptic physiology and the functional
parameters that are relevant to the intact
system will come only from studying singlesynapse physiology in vivo, which at the
present time is still a formidable challenge.

Conclusion and future directions

Heterogeneity in the synaptic weights of a
connection seems to result, in part, from
individualized regulation of pr , and as we
discussed above it has important functional

5.

380 | may 2009 | volUmE 10

Tiago Branco is at the Wolfson Institute for Biomedical

Research and Department of Neuroscience, Physiology
and Pharmacology, University College London, WC1E
6BT, UK.
Kevin Staras is at the School of Life Sciences, University
of Sussex, BN1 9QG, UK.
e-mails: t.branco@ucl.ac.uk; k.staras@sussex.ac.uk
doi:10.1038/nrn2634
1.
2.
3.

4.

6.

Del Castillo, J. & Katz, B. Quantal components of the

end-plate potential. J. Physiol. 124, 560573
(1954).
Sudhof, T. C. The synaptic vesicle cycle. Annu. Rev.
Neurosci. 27, 509547 (2004).
Tracey, D. J. & Walmsley, B. Synaptic input from
identified muscle afferents to neurones of the dorsal
spinocerebellar tract in the cat. J. Physiol. 350,
599614 (1984).
Korn, H., Mallet, A., Triller, A. & Faber, D. S.
Transmission at a central inhibitory synapse. II.
Quantal description of release, with a physical
correlate for binomial n. J. Neurophysiol. 48,
679707 (1982).
Korn, H., Triller, A., Mallet, A. & Faber, D. S.
Fluctuating responses at a central synapse: n of
binomial fit predicts number of stained presynaptic
boutons. Science 213, 898901 (1981).
Harvey, R. J. & Napper, R. M. Quantitative study of
granule and Purkinje cells in the cerebellar cortex of
the rat. J. Comp. Neurol. 274, 151157 (1988).

www.nature.com/reviews/neuro
2009 Macmillan Publishers Limited. All rights reserved

PersPectives
7.

8.

9.

10.

11.
12.
13.

14.

15.

16.
17.
18.
19.
20.
21.

22.

23.

24.

25.

26.
27.

28.

29.

30.

31.

Deuchars, J., West, D. C. & Thomson, A. M.

Relationships between morphology and physiology of
pyramid-pyramid single axon connections in rat
neocortex in vitro. J. Physiol. 478, 423435 (1994).
Feldmeyer, D., Lubke, J. & Sakmann, B. Efficacy and
connectivity of intracolumnar pairs of layer 2/3
pyramidal cells in the barrel cortex of juvenile rats.
J. Physiol. 575, 583602 (2006).
Buhl, E. H., Halasy, K. & Somogyi, P. Diverse sources
of hippocampal unitary inhibitory postsynaptic
potentials and the number of synaptic release sites.
Nature 368, 823828 (1994).
Tamas, G., Buhl, E. H. & Somogyi, P. Fast IPSPs
elicited via multiple synaptic release sites by different
types of GABAergic neurone in the cat visual cortex.
J. Physiol. 500, 715738 (1997).
Llinas, R., Bloedel, J. R. & Hillman, D. E. Functional
characterization of neuronal circuitry of frog cerebellar
cortex. J. Neurophysiol. 32, 847870 (1969).
Isaacson, J. S. & Walmsley, B. Counting quanta: direct
measurements of transmitter release at a central
synapse. Neuron 15, 875884 (1995).
Meyer, A. C., Neher, E. & Schneggenburger, R.
Estimation of quantal size and number of functional
active zones at the calyx of held synapse by
nonstationary EPSC variance analysis. J. Neurosci. 21,
78897900 (2001).
Branco, T., Staras, K., Darcy, K. J. & Goda, Y. Local
dendritic activity sets release probability at
hippocampal synapses. Neuron 59, 475485
(2008).
Buhl, E. H. et al. Effect, number and location of
synapses made by single pyramidal cells onto aspiny
interneurones of cat visual cortex. J. Physiol. 500,
689713 (1997).
Thomson, A. M. & Lamy, C. Functional maps of
neocortical local circuitry. Front. Neurosci. 1, 1942
(2007).
Rall, W. in Neural Theory and Modeling (ed. Reiss, R. F.)
(Stanford Univ. Press, Palo Alto, 1964).
Polsky, A., Mel, B. W. & Schiller, J. Computational
subunits in thin dendrites of pyramidal cells. Nature
Neurosci. 7, 621627 (2004).
Katz, B. (ed.) Nerve, Muscle and Synapse (McGrawHill, 1966).
Maass, W. & Zador, A. M. Dynamic stochastic
synapses as computational units. Neural Comput. 11,
903917 (1999).
Jack, J. J., Redman, S. J. & Wong, K. The components
of synaptic potentials evoked in cat spinal
motoneurones by impulses in single group Ia afferents.
J. Physiol. 321, 6596 (1981).
Redman, S. & Walmsley, B. Amplitude fluctuations in
synaptic potentials evoked in cat spinal motoneurones
at identified group Ia synapses. J. Physiol. 343,
135145 (1983).
Walmsley, B., Edwards, F. R. & Tracey, D. J. Nonuniform
release probabilities underlie quantal synaptic
transmission at a mammalian excitatory central
synapse. J. Neurophysiol. 60, 889908 (1988).
Clamann, H. P., Mathis, J. & Luscher, H. R. Variance
analysis of excitatory postsynaptic potentials in cat
spinal motoneurons during posttetanic potentiation.
J. Neurophysiol. 61, 403416 (1989).
Silver, R. A., Momiyama, A. & Cull-Candy, S. G. Locus
of frequency-dependent depression identified with
multiple-probability fluctuation analysis at rat climbing
fibre-Purkinje cell synapses. J. Physiol. 510, 881902
(1998).
Hessler, N. A., Shirke, A. M. & Malinow, R. The
probability of transmitter release at a mammalian
central synapse. Nature 366, 569572 (1993).
Rosenmund, C., Clements, J. D. & Westbrook, G. L.
Nonuniform probability of glutamate release at a
hippocampal synapse. Science 262, 754757
(1993).
Murthy, V. N., Sejnowski, T. J. & Stevens, C. F.
Heterogeneous release properties of visualized
individual hippocampal synapses. Neuron 18,
599612 (1997).
Slutsky, I., Sadeghpour, S., Li, B. & Liu, G.
Enhancement of synaptic plasticity through chronically
reduced Ca2+ flux during uncorrelated activity. Neuron
44, 835849 (2004).
Granseth, B., Odermatt, B., Royle, S. J. & Lagnado, L.
Clathrin-mediated endocytosis is the dominant
mechanism of vesicle retrieval at hippocampal
synapses. Neuron 51, 773786 (2006).
Schikorski, T. & Stevens, C. F. Quantitative
ultrastructural analysis of hippocampal excitatory
synapses. J. Neurosci. 17, 58585867 (1997).

32. Bittner, G. D. Differentiation of nerve terminals in the

crayfish opener muscle and its functional significance.
J. Gen. Physiol. 51, 731758 (1968).
33. Atwood, H. L. & Bittner, G. D. Matching of excitatory
and inhibitory inputs to crustacean muscle fibers.
J. Neurophysiol. 34, 157170 (1971).
34. Frank, E. Matching of facilitation at the neuromuscular
junction of the lobster: a possible case for influence of
muscle on nerve. J. Physiol. 233, 635658 (1973).
35. Cooper, R. L., Harrington, C. C., Marin, L. & Atwood,
H. L. Quantal release at visualized terminals of a
crayfish motor axon: intraterminal and regional
differences. J. Comp. Neurol. 375, 583600 (1996).
36. Parnas, I. Differential block at high frequency of
branches of a single axon innervating two muscles.
J. Neurophysiol. 35, 903914 (1972).
37. Bennett, M. R., Jones, P. & Lavidis, N. A. The
probability of quantal secretion along visualized
terminal branches at amphibian (Bufo marinus)
neuromuscular synapses. J. Physiol. 379, 257274
(1986).
38. Robitaille, R. & Tremblay, J. P. Non-uniform release at
the frog neuromuscular junction: evidence of
morphological and physiological plasticity. Brain Res.
434, 95116 (1987).
39. Robitaille, R. & Tremblay, J. P. Non-uniform responses
to Ca2+ along the frog neuromuscular junction: effects
on the probability of spontaneous and evoked
transmitter release. Neuroscience 40, 571585
(1991).
40. Katz, P. S., Kirk, M. D. & Govind, C. K. Facilitation and
depression at different branches of the same motor
axon: evidence for presynaptic differences in release.
J. Neurosci. 13, 30753089 (1993).
41. Muller, K. J. & Nicholls, J. G. Different properties of
synapses between a single sensory neurone and two
different motor cells in the leech C.N.S. J. Physiol.
238, 357369 (1974).
42. Koerber, H. R. & Mendell, L. M. Modulation of
synaptic transmission at Ia-afferent fiber connections
on motoneurons during high-frequency stimulation:
role of postsynaptic target. J. Neurophysiol. 65,
590597 (1991).
43. Brodin, L., Shupliakov, O., Pieribone, V. A., Hellgren, J.
& Hill, R. H. The reticulospinal glutamate synapse in
lamprey: plasticity and presynaptic variability.
J. Neurophysiol. 72, 592604 (1994).
44. Mennerick, S. & Zorumski, C. F. Paired-pulse
modulation of fast excitatory synaptic currents in
microcultures of rat hippocampal neurons. J. Physiol.
488, 85101 (1995).
45. Markram, H., Wang, Y. & Tsodyks, M. Differential
signaling via the same axon of neocortical pyramidal
neurons. Proc. Natl Acad. Sci. USA 95, 53235328
(1998).
46. Reyes, A. et al. Target-cell-specific facilitation and
depression in neocortical circuits. Nature Neurosci. 1,
279285 (1998).
47. Davis, G. W. & Murphey, R. K. A role for postsynaptic
neurons in determining presynaptic release properties
in the cricket CNS: evidence for retrograde control of
facilitation. J. Neurosci. 13, 38273838 (1993).
48. Laurent, G. & Sivaramakrishnan, A. Single local
interneurons in the locust make central synapses with
different properties of transmitter release on distinct
postsynaptic neurons. J. Neurosci. 12, 23702380
(1992).
49. Koester, H. J. & Johnston, D. Target cell-dependent
normalization of transmitter release at neocortical
synapses. Science 308, 863866 (2005).
50. Zucker, R. S. & Regehr, W. G. Short-term synaptic
plasticity. Annu. Rev. Physiol. 64, 355405 (2002).
51. Waters, J. & Smith, S. J. Vesicle pool partitioning
influences presynaptic diversity and weighting in rat
hippocampal synapses. J. Physiol. 541, 811823
(2002).
52. He, L. & Wu, L. G. The debate on the kiss-and-run
fusion at synapses. Trends Neurosci. 30, 447455
(2007).
53. Catterall, W. A. & Few, A. P. Calcium channel
regulation and presynaptic plasticity. Neuron 59,
882901 (2008).
54. Borst, J. G. & Sakmann, B. Effect of changes in action
potential shape on calcium currents and transmitter
release in a calyx-type synapse of the rat auditory
brainstem. Philos. Trans. R. Soc. Lond. B Biol. Sci.
354, 347355 (1999).
55. Meinrenken, C. J., Borst, J. G. G. & Sakmann, B. Local
routes revisited: the space and time dependence of
the Ca2+ signal for phasic transmitter release at the
rat calyx of Held. J. Physiol. 547, 665689 (2003).

naTURE REvIEWS | NeuroscieNce

56. Augustine, G. J., Santamaria, F. & Tanaka, K. Local

calcium signaling in neurons. Neuron 40, 331346
(2003).
57. Schneggenburger, R. & Neher, E. Presynaptic calcium
and control of vesicle fusion. Curr. Opin. Neurobiol.
15, 266274 (2005).
58. Reid, C. A., Clements, J. D. & Bekkers, J. M.
Nonuniform distribution of Ca2+ channel subtypes on
presynaptic terminals of excitatory synapses in
hippocampal cultures. J. Neurosci. 17, 27382745
(1997).
59. Reuter, H. Measurements of exocytosis from single
presynaptic nerve terminals reveal heterogeneous
inhibition by Ca2+-channel blockers. Neuron 14,
773779 (1995).
60. Sorra, K. E. & Harris, K. M. Occurrence and threedimensional structure of multiple synapses between
individual radiatum axons and their target pyramidal
cells in hippocampal area CA1. J. Neurosci. 13,
37363748 (1993).
61. Shigemoto, R. et al. Target-cell-specific concentration
of a metabotropic glutamate receptor in the
presynaptic active zone. Nature 381, 523525
(1996).
62. Scanziani, M., Gahwiler, B. H. & Charpak, S. Target
cell-specific modulation of transmitter release at
terminals from a single axon. Proc. Natl Acad. Sci.
USA 95, 1200412009 (1998).
63. Nicoll, R. A. & Schmitz, D. Synaptic plasticity at
hippocampal mossy fibre synapses. Nature Rev.
Neurosci. 6, 863876 (2005).
64. Maccaferri, G., Toth, K. & McBain, C. J. Target-specific
expression of presynaptic mossy fiber plasticity.
Science 279, 13681370 (1998).
65. Pelkey, K. A. & McBain, C. J. Target-cell-dependent
plasticity within the mossy fibre-CA3 circuit reveals
compartmentalized regulation of presynaptic function
at divergent release sites. J. Physiol. 586,
14951502 (2008).
66. Lei, S. & McBain, C. J. Distinct NMDA receptors provide
differential modes of transmission at mossy fiberinterneuron synapses. Neuron 33, 921933 (2002).
67. Chevaleyre, V. & Castillo, P. E. Heterosynaptic LTD of
hippocampal GABAergic synapses: a novel role of
endocannabinoids in regulating excitability. Neuron
38, 461472 (2003).
68. Gerdeman, G. L., Ronesi, J. & Lovinger, D. M.
Postsynaptic endocannabinoid release is critical to
long-term depression in the striatum. Nature
Neurosci. 5, 446451 (2002).
69. Marsicano, G. et al. The endogenous cannabinoid
system controls extinction of aversive memories.
Nature 418, 530534 (2002).
70. Robbe, D., Kopf, M., Remaury, A., Bockaert, J. &
Manzoni, O. J. Endogenous cannabinoids mediate
long-term synaptic depression in the nucleus
accumbens. Proc. Natl Acad. Sci. USA 99,
83848388 (2002).
71. Safo, P. K. & Regehr, W. G. Endocannabinoids control
the induction of cerebellar LTD. Neuron 48, 647659
(2005).
72. Sjostrom, P. J., Turrigiano, G. G. & Nelson, S. B.
Neocortical LTD via coincident activation of
presynaptic NMDA and cannabinoid receptors.
Neuron 39, 641654 (2003).
73. Chevaleyre, V., Takahashi, K. A. & Castillo, P. E.
Endocannabinoid-mediated synaptic plasticity in the
CNS. Annu. Rev. Neurosci. 29, 3776 (2006).
74. Davis, G. W. & Goodman, C. S. Synapse-specific
control of synaptic efficacy at the terminals of a single
neuron. Nature 392, 8286 (1998).
75. Stewart, B. A., Schuster, C. M., Goodman, C. S. &
Atwood, H. L. Homeostasis of synaptic transmission in
Drosophila with genetically altered nerve terminal
morphology. J. Neurosci. 16, 38773886 (1996).
76. Paradis, S., Sweeney, S. T. & Davis, G. W. Homeostatic
control of presynaptic release is triggered by
postsynaptic membrane depolarization. Neuron 30,
737749 (2001).
77. Petersen, S. A., Fetter, R. D., Noordermeer, J. N.,
Goodman, C. S. & DiAntonio, A. Genetic analysis of
glutamate receptors in Drosophila reveals a
retrograde signal regulating presynaptic transmitter
release. Neuron 19, 12371248 (1997).
78. Sandrock, A. W. J. et al. Maintenance of acetylcholine
receptor number by neuregulins at the neuromuscular
junction in vivo. Science 276, 599603 (1997).
79. Frank, C. A., Kennedy, M. J., Goold, C. P., Marek, K. W.
& Davis, G. W. Mechanisms underlying the rapid
induction and sustained expression of synaptic
homeostasis. Neuron 52, 663677 (2006).

volUmE 10 | may 2009 | 381

2009 Macmillan Publishers Limited. All rights reserved

PersPectives
80. Frank, C. A., Pielage, J. & Davis, G. W. A presynaptic
homeostatic signaling system composed of the Eph
receptor, ephexin, Cdc42, and CaV2.1 calcium
channels. Neuron 61, 556569 (2009).
81. Haghighi, A. P. et al. Retrograde control of synaptic
transmission by postsynaptic CaMKII at the
Drosophila neuromuscular junction. Neuron 39,
255267 (2003).
82. Bacci, A. et al. Chronic blockade of glutamate
receptors enhances presynaptic release and
downregulates the interaction between synaptophysinsynaptobrevin-vesicle-associated membrane protein 2.
J. Neurosci. 21, 65886596 (2001).
83. Thiagarajan, T. C., Lindskog, M. & Tsien, R. W.
Adaptation to synaptic inactivity in hippocampal
neurons. Neuron 47, 725737 (2005).
84. Thiagarajan, T. C., Piedras-Renteria, E. S. & Tsien,
R. W. - and CaMKII. Inverse regulation by neuronal
activity and opposing effects on synaptic strength.
Neuron 36, 11031114 (2002).
85. Murthy, V. N., Schikorski, T., Stevens, C. F. & Zhu, Y.
Inactivity produces increases in neurotransmitter
release and synapse size. Neuron 32, 673682
(2001).
86. Darcy, K. J., Staras, K., Collinson, L. M. & Goda, Y.
Constitutive sharing of recycling synaptic vesicles
between presynaptic boutons. Nature Neurosci. 9,
315321 (2006).
87. Staras, K. Share and share alike: trading of
presynaptic elements between central synapses.
Trends Neurosci. 30, 292298 (2007).
88. Galante, M., Avossa, D., Rosato-Siri, M. & Ballerini, L.
Homeostatic plasticity induced by chronic block of
AMPA/kainate receptors modulates the generation of
rhythmic bursting in rat spinal cord organotypic
cultures. Eur. J. Neurosci. 14, 903917 (2001).
89. Wierenga, C. J., Walsh, M. F. & Turrigiano, G. G.
Temporal regulation of the expression locus of
homeostatic plasticity. J. Neurophysiol. 96,
21272133 (2006).
90. Hardingham, N. R., Hardingham, G. E., Fox, K. D. &
Jack, J. J. B. Presynaptic efficacy directs normalization
of synaptic strength in layer 2/3 rat neocortex after
paired activity. J. Neurophysiol. 97, 29652975
(2007).
91. Goold, C. P. & Davis, G. W. The BMP ligand Gbb gates
the expression of synaptic homeostasis independent
of synaptic growth control. Neuron 56, 109123
(2007).
92. Futai, K. et al. Retrograde modulation of presynaptic
release probability through signaling mediated by
PSD-95-neuroligin. Nature Neurosci. 10, 186195
(2007).
93. Alger, B. E. Retrograde signaling in the regulation of
synaptic transmission: focus on endocannabinoids.
Prog. Neurobiol. 68, 247286 (2002).
94. Wilson, R. I., Kunos, G. & Nicoll, R. A. Presynaptic
specificity of endocannabinoid signaling in the
hippocampus. Neuron 31, 453462 (2001).
95. Wilson, R. I. & Nicoll, R. A. Endogenous cannabinoids
mediate retrograde signalling at hippocampal
synapses. Nature 410, 588592 (2001).
96. Beierlein, M., Fioravante, D. & Regehr, W. G.
Differential expression of posttetanic potentiation and
retrograde signaling mediate target-dependent shortterm synaptic plasticity. Neuron 54, 949959 (2007).
97. Brasier, D. J. & Feldman, D. E. Synapse-specific
expression of functional presynaptic NMDA receptors
in rat somatosensory cortex. J. Neurosci. 28,
21992211 (2008).
98. Brenowitz, S. D. & Regehr, W. G. Reliability and
heterogeneity of calcium signaling at single
presynaptic boutons of cerebellar granule cells.
J. Neurosci. 27, 78887898 (2007).
99. Moulder, K. L. et al. Vesicle pool heterogeneity at
hippocampal glutamate and GABA synapses.
J. Neurosci. 27, 98469854 (2007).
100. Sun, H. Y., Lyons, S. A. & Dobrunz, L. E. Mechanisms
of target-cell specific short-term plasticity at Schaffer
collateral synapses onto interneurones versus
pyramidal cells in juvenile rats. J. Physiol. 568,
815840 (2005).
101. Wyatt, R. M. & Balice-Gordon, R. J. Activity-dependent
elimination of neuromuscular synapses. J. Neurocytol.
32, 777794 (2003).
102. Atwood, H. L. & Karunanithi, S. Diversification of
synaptic strength: presynaptic elements. Nature Rev.
Neurosci. 3, 497516 (2002).
103. Craig, A. M. & Boudin, H. Molecular heterogeneity of
central synapses: afferent and target regulation.
Nature Neurosci. 4, 569578 (2001).

104. Tyler, W. J. et al. BDNF increases release probability

and the size of a rapidly recycling vesicle pool within
rat hippocampal excitatory synapses. J. Physiol. 574,
787803 (2006).
105. Turrigiano, G. G. The self-tuning neuron: synaptic
scaling of excitatory synapses. Cell 135, 422435
(2008).
106. Abbott, L. F. & Regehr, W. G. Synaptic computation.
Nature 431, 796803 (2004).
107. Goldman, M. S. Enhancement of information
transmission efficiency by synaptic failures. Neural
Comput. 16, 11371162 (2004).
108. Moore, E. F. & Shannon, C. E. Reliable circuits using
less reliable relays. J. Franklin. Inst. 262, 191208,
281297 (1956).
109. Smetters, D. K. & Zador, A. Synaptic transmission:
noisy synapses and noisy neurons. Curr. Biol. 6,
12171218 (1996).
110. Stevens, C. F. Neuronal communication. Cooperativity
of unreliable neurons. Curr. Biol. 4, 268269
(1994).
111. Zador, A. Impact of synaptic unreliability on the
information transmitted by spiking neurons.
J. Neurophysiol. 79, 12191229 (1998).
112. Lisman, J. & Raghavachari, S. A unified model of the
presynaptic and postsynaptic changes during LTP at
CA1 synapses. Sci. STKE 2006, re11 (2006).
113. Thomson, A. M. Presynaptic frequency- and patterndependent filtering. J. Comput. Neurosci. 15,
159202 (2003).
114. Tsodyks, M. V. & Markram, H. The neural code
between neocortical pyramidal neurons depends on
neurotransmitter release probability. Proc. Natl Acad.
Sci. USA 94, 719723 (1997).
115. Stevens, C. F. & Wang, Y. Changes in reliability of
synaptic function as a mechanism for plasticity. Nature
371, 704707 (1994).
116. Manwani, A. & Koch, C. Detecting and estimating
signals over noisy and unreliable synapses:
information-theoretic analysis. Neural Comput. 13,
133 (2001).
117. Fernandez-Alfonso, T. & Ryan, T. A. The efficiency of
the synaptic vesicle cycle at central nervous system
synapses. Trends Cell Biol. 16, 413420 (2006).
118. Saviane, C. & Silver, R. A. Fast vesicle reloading and a
large pool sustain high bandwidth transmission at a
central synapse. Nature 439, 983987 (2006).
119. Zhang, Q., Li, Y. & Tsien, R. W. The dynamic control of
kiss-and-run and vesicular reuse probed with single
nanoparticles. Science 323, 14481453 (2009).
120. Attwell, D. & Gibb, A. Neuroenergetics and the kinetic
design of excitatory synapses. Nature Rev. Neurosci.
6, 841849 (2005).
121. Attwell, D. & Laughlin, S. B. An energy budget for
signaling in the grey matter of the brain. J. Cereb.
Blood Flow Metab. 21, 11331145 (2001).
122. Laughlin, S. B., de Ruyter van Steveninck, R. R. &
Anderson, J. C. The metabolic cost of neural
information. Nature Neurosci. 1, 3641 (1998).
123. Levy, W. B. & Baxter, R. A. Energy efficient neural
codes. Neural Comput. 8, 531543 (1996).
124. Levy, W. B. & Baxter, R. A. Energy-efficient neuronal
computation via quantal synaptic failures. J. Neurosci.
22, 47464755 (2002).
125. Rabinowitch, I. & Segev, I. The interplay between
homeostatic synaptic plasticity and functional
dendritic compartments. J. Neurophysiol. 96,
276283 (2006).
126. Rall, W. & Rinzel, J. Branch input resistance and
steady attenuation for input to one branch of a
dendritic neuron model. Biophys. J. 13, 648687
(1973).
127. Rinzel, J. & Rall, W. Transient response in a dendritic
neuron model for current injected at one branch.
Biophys. J. 14, 759790 (1974).
128. Johnston, D. & Narayanan, R. Active dendrites:
colorful wings of the mysterious butterflies. Trends
Neurosci. 31, 309316 (2008).
129. Rall, W. Distinguishing theoretical synaptic potentials
computed for different soma-dendritic distributions of
synaptic input. J. Neurophysiol. 30, 11381168
(1967).
130. Mel, B. W. Synaptic integration in an excitable
dendritic tree. J. Neurophysiol. 70, 10861101
(1993).
131. Poirazi, P., Brannon, T. & Mel, B. W. Pyramidal neuron
as two-layer neural network. Neuron 37, 989999
(2003).
132. Stellwagen, D. & Malenka, R. C. Synaptic scaling
mediated by glial TNF-. Nature 440, 10541059
(2006).

382 | may 2009 | volUmE 10

133. Redman, S. & Faber, D. S. Editorial. J. Neurosci.

Methods 130, 103104 (2003).
134. Denk, W. & Horstmann, H. Serial block-face
scanning electron microscopy to reconstruct threedimensional tissue nanostructure. PLoS Biol. 2, e329
(2004).
135. Luo, L., Callaway, E. M. & Svoboda, K. Genetic
dissection of neural circuits. Neuron 57, 634660
(2008).
136. Dobrunz, L. E. & Stevens, C. F. Heterogeneity of
release probability, facilitation, and depletion at
central synapses. Neuron 18, 9951008 (1997).
137. Feldmeyer, D. & Sakmann, B. Synaptic efficacy and
reliability of excitatory connections between the
principal neurones of the input (layer 4) and output
layer (layer 5) of the neocortex. J. Physiol. 525,
3139 (2000).
138. Markram, H., Lubke, J., Frotscher, M., Roth, A. &
Sakmann, B. Physiology and anatomy of synaptic
connections between thick tufted pyramidal neurones
in the developing rat neocortex. J. Physiol. 500,
409440 (1997).
139. Chen, G., Harata, N. C. & Tsien, R. W. Paired-pulse
depression of unitary quantal amplitude at single
hippocampal synapses. Proc. Natl Acad. Sci. USA 101,
10631068 (2004).
140. Oertner, T. G., Sabatini, B. L., Nimchinsky, E. A. &
Svoboda, K. Facilitation at single synapses probed
with optical quantal analysis. Nature Neurosci. 5,
657664 (2002).
141. Zakharenko, S. S., Zablow, L. & Siegelbaum, S. A.
Visualization of changes in presynaptic function during
long-term synaptic plasticity. Nature Neurosci. 4,
711717 (2001).
142. Gandhi, S. P. & Stevens, C. F. Three modes of synaptic
vesicular recycling revealed by single-vesicle imaging.
Nature 423, 607613 (2003).
143. Balaji, J. & Ryan, T. A. Single-vesicle imaging reveals
that synaptic vesicle exocytosis and endocytosis are
coupled by a single stochastic mode. Proc. Natl Acad.
Sci. USA 104, 2057620581 (2007).
144. Emptage, N. J., Reid, C. A., Fine, A. & Bliss, T. V. P.
Optical quantal analysis reveals a presynaptic
component of LTP at hippocampal Schafferassociational synapses. Neuron 38, 797804
(2003).
145. Yuste, R., Majewska, A., Cash, S. S. & Denk, W.
Mechanisms of calcium influx into hippocampal
spines: heterogeneity among spines, coincidence
detection by NMDA receptors, and optical quantal
analysis. J. Neurosci. 19, 19761987 (1999).
146. Brown, A. G. & Fyffe, R. E. Direct observations on the
contacts made between Ia afferent fibres and alphamotoneurones in the cats lumbosacral spinal cord.
J. Physiol. 313, 121140 (1981).
147. Grantyn, R., Shapovalov, A. I. & Shiriaev, B. I. Relation
between structural and release parameters at the frog
sensory-motor synapse. J. Physiol. 349, 459474
(1984).
148. Hardingham, N. R. et al. Extracellular calcium
regulates postsynaptic efficacy through group 1
metabotropic glutamate receptors. J. Neurosci. 26,
63376345 (2006).
149. Bremaud, A., West, D. C. & Thomson, A. M. Binomial
parameters differ across neocortical layers and with
different classes of connections in adult rat and cat
neocortex. Proc. Natl Acad. Sci. USA 104,
1413414139 (2007).
150. Feldmeyer, D., Lubke, J., Silver, R. A. & Sakmann, B.
Synaptic connections between layer 4 spiny neuronelayer 2/3 pyramidal cell pairs in juvenile rat barrel
cortex: physiology and anatomy of interlaminar
signalling within a cortical column. J. Physiol. 538,
803822 (2002).
151. Silver, R. A., Lubke, J., Sakmann, B. & Feldmeyer, D.
High-probability uniquantal transmission at excitatory
synapses in barrel cortex. Science 302, 19811984
(2003).
152. Feldmeyer, D., Egger, V., Lubke, J. & Sakmann, B.
Reliable synaptic connections between pairs of
excitatory layer 4 neurones within a single barrel of
developing rat somatosensory cortex. J. Physiol. 521,
169190 (1999).
153. Tarczy-Hornoch, K., Martin, K. A., Stratford, K. J. &
Jack, J. J. Intracortical excitation of spiny neurons in
layer 4 of cat striate cortex in vitro. Cereb. Cortex 9,
833843 (1999).
154. Thomson, A. M., Deuchars, J. & West, D. C.
Neocortical local synaptic circuitry revealed with dual
intracellular recordings and biocytin-filling. J. Physiol.
(Paris) 90, 211215 (1996).

www.nature.com/reviews/neuro
2009 Macmillan Publishers Limited. All rights reserved

PersPectives
155. Deuchars, J. & Thomson, A. M. Innervation of burst
firing spiny interneurons by pyramidal cells in deep
layers of rat somatomotor cortex: paired intracellular
recordings with biocytin filling. Neuroscience 69,
739755 (1995).
156. Deuchars, J. & Thomson, A. M. CA1 pyramid-pyramid
connections in rat hippocampus in vitro: dual
intracellular recordings with biocytin filling.
Neuroscience 74, 10091018 (1996).
157. Larkman, A. U., Jack, J. J. & Stratford, K. J. Quantal
analysis of excitatory synapses in rat hippocampal
CA1 in vitro during low-frequency depression.
J. Physiol. 505, 457471 (1997).
158. Biro, A. A., Holderith, N. B. & Nusser, Z. Release
probability-dependent scaling of the postsynaptic
responses at single hippocampal GABAergic synapses.
J. Neurosci. 26, 1248712496 (2006).
159. Gulyas, A. I. et al. Hippocampal pyramidal cells excite
inhibitory neurons through a single release site.
Nature 366, 683687 (1993).
160. Dittman, J. S., Kreitzer, A. C. & Regehr, W. G. Interplay
between facilitation, depression, and residual calcium
at three presynaptic terminals. J. Neurosci. 20,
13741385 (2000).
161. Auger, C., Kondo, S. & Marty, A. Multivesicular release
at single functional synaptic sites in cerebellar stellate
and basket cells. J. Neurosci. 18, 45324547 (1998).
162. Ding, J., Peterson, J. D. & Surmeier, D. J. Corticostriatal
and thalamostriatal synapses have distinctive
properties. J. Neurosci. 28, 64836492 (2008).

163. Sahara, Y. & Takahashi, T. Quantal components of

the excitatory postsynaptic currents at a rat central
auditory synapse. J. Physiol. 536, 189197 (2001).
164. Sakaba, T., Schneggenburger, R. & Neher, E.
Estimation of quantal parameters at the calyx of
Held synapse. Neurosci. Res. 44, 343356 (2002).
165. Murphy, G. J., Glickfeld, L. L., Balsen, Z. &
Isaacson, J. S. Sensory neuron signaling to the
brain: properties of transmitter release from
olfactory nerve terminals. J. Neurosci. 24,
30233030 (2004).

Acknowledgements

We would like to thank D. Attwell, B. Clark, A. Roth and M.

Husser for critical comments on the manuscript, M. London
for helpful discussions, and Y. Goda and M. Husser for support. T.B. is supported by grants from the Wellcome Trust and
the Gatsby Charitable Foundation (to M. Husser). K.S. is
supported by the Wellcome Trust (WT084357MF) and the
Biotechnology and Biological Sciences Research Council (BB/
F018371). We apologise to authors whose work we were
unable to include owing to space constraints.

FURTHER inFORMATiOn
Kevin starass homepage: http://www.sussex.ac.uk/
biology/profile16600.html
All liNks Are ActiVe iN the oNliNe PDF

OpiniOn

Phasic acetylcholine release and

the volume transmission hypothesis:
time to move on
Martin Sarter, Vinay Parikh and W. Matthew Howe

Abstract | Traditional descriptions of the cortical cholinergic input system focused

on the diffuse organization of cholinergic projections and the hypothesis that
slowly changing levels of extracellular acetylcholine (Ach) mediate different
arousal states. The ability of Ach to reach the extrasynaptic space (volume
neurotransmission), as opposed to remaining confined to the synaptic cleft
(wired neurotransmission), has been considered an integral component of this
conceptualization. recent studies demonstrated that phasic release of Ach, at
the scale of seconds, mediates precisely defined cognitive operations. This
characteristic of cholinergic neurotransmission is proposed to be of primary
importance for understanding cholinergic function and developing treatments for
cognitive disorders that result from abnormal cholinergic neurotransmission.
The entire cortical mantle is innervated by
cholinergic neurons that originate in the
nucleus basalis of meynert, the substantia
innominata and the horizontal limb of the
diagonal band all structures of the basal
forebrain (BF) (FIG. 1). Traditionally, the cortical cholinergic input system has been categorized as the rostral component of the brains
ascending arousal systems, complementing
the modulatory roles of, and interacting
with, noradrenergic, serotonergic and other
projection systems that broadly influence the

readiness of the forebrain for input processing, wakefulness and somnolence1. However,
more recent evidence has supported the more
specific hypothesis that cortical cholinergic
inputs mediate essential aspects of attentional
information processing29. as a result, efforts
to develop treatments for a wide range of
cognitive disorders have focused on cholinomimetic approaches, particularly acetylcholinesterase (aCHE) inhibitors and agonists at
muscarinic (m) and nicotinic (n) acetylcholine
(aCh) receptors (aChRs)1012.

naTURE REvIEWS | NeuroscieNce

The anatomical organization of the cortical cholinergic input system seems to be

largely consistent with the notion of a diffuse
pathway (this article does not address the
hippocampal cholinergic projection system
or cholinergic projections to the amygdala).
Tracing studies revealed a roughly ventrolateral, dorsomedial and rostrocaudal
topographical organization of cholinergic
BF projections but did not suggest a more
precise topography that would indicate, for
example, that adjacent neurons in the BF
innervate adjacent regions in the cortex1316
(FIG. 1b,c). nearly all cortical layers and regions
are innervated by BF cholinergic neurons17,
although the distribution of choline acetyltransferase (CHaT)- or aCHE-positive
fibres in the cortex indicates differences in
the density of the cholinergic innervation of
specific layers1821 (FIG. 2). This seemingly diffuse organization of the cortical cholinergic
input system has supported descriptions that
it exerts general, uniform effects across the
cortical hemispheres20.
In contrast to other diffusely organized
ascending systems, such as the ascending reticular systems of the brainstem, the
axons of corticopetal cholinergic neurons
(subcortical afferents that project to both
cerebral hemispheres) do not seem to be
extensively collateralized: individual neurons innervate a relatively small cortical
field2224. Thus, separate cortical regions,
such as frontal and parietal regions, are
not innervated by the same cholinergic
neurons, suggesting that these regions
may be differentially modulated by the
cholinergic input system.
It has recently been proposed14,15,25 that
the corticopetal cholinergic system is less
diffusely organized than was traditionally assumed (FIG. 1b,c). In support of this
hypothesis, it has been demonstrated that
there are clusters of cholinergic cells in the
BF15,25,26 and that the BF receives modalityspecific projections27. The morphological
heterogeneity of BF cholinergic neurons (see
rEFs 28,29) and of their efferent and afferent
projection systems, including the degree to
which they exhibit a topographical organization, remains insufficiently understood13.
For example, the finding that manipulations
of the excitability of the nucleus accumbens
affect prefrontal aCh release but not the
release of aCh in parietal regions30,31 does
not correspond with traditional descriptions
of the organization of this system: it is more
consistent with views suggesting a refined
anatomical or functional topographical
organization of the BF corticopetal
projection system.
volUmE 10 | may 2009 | 383

2009 Macmillan Publishers Limited. All rights reserved